How the quality of call detail records influences the detection of commuting trips



Call Detail Records provide information on the origin and destination of voice calls at the level of the base stations in a cellular network. The low spatial resolution and sparsity of these data constitutes challenges in using them for mo- bility characterization. In this paper we analyze the impact on the detection of commuting patterns of four parameters: density of base stations per square kilo- meter, average number of calls made and received per day per user, regularity of these calls, and the number of active days per user. In this study, we use CDRs collected from Portugal over a period of fourteen months. Based on the result of our study, we are able to infer the commuting patterns of 10.42% of the users in our data set by considering users with at least 7.5 calls per day. Accounting users with over 7.5 calls per day, on average, does not result in a significant improve- ment on the result. Concerning the inference of routes in the home-to-work di- rection and vice versa, we examined users who connect to the cellular network, on average, every 17 days to everyday, which results in a 0.27% to 11.1% of trips detected, respectively. Finally, we found that with 208 days of data we are able to infer 5.67% of commuting trips and this percentage does not improve signifi- cantly by considering more data.


Call Detail Records, Commuting Trip, Origin-Destination Matrices


Analysis of CDRs

Related Project



EPIA Conference on Artificial Intelligence 2019, July 2019


Cited by

No citations found